TL;DR

Clinical trial analysis is still shockingly manual.

After a study collects data, teams spend months converting raw clinical data into SDTM, ADaM, TFLs, and submission-ready packages. This work is slow, expensive, and error-prone.

Astraea is a platform that completely automates the end-to-end biometrics pipeline from just a protocol and the raw data to a submission ready output. We are starting with oncology and rare disease studies, where analysis complexity is high and every month saved matters.

The Problem

Today, sponsors and CROs still rely on large teams of statistical programmers, biostatisticians, and data managers to manually transform messy clinical data into regulatory outputs.

A single study can require:

1. Raw data cleaning and reconciliation
2. SDTM mapping
3. ADaM dataset creation
4. TFL shell generation
5. Statistical programming
6. QC and validation
7. Regulatory formatting and traceability

From our customer conversations, this can take around 9 months with teams of 5-10 people communicating asynchronously. And the process is still full of Word docs, Excel specs, SAS programs, manual handoffs, version-control issues, and repeated QC cycles.

Ultimately, this means slower decisions, delayed filings, higher CRO spend, and less control over the most important data in the company.

Our Solution

We built the first AI native platform for clinical trial biometrics.

Astraea takes study inputs and automatically generates the datasets, analysis specs, statistical code, and outputs required for regulatory review.

The pipeline:

1. Ingest raw clinical data, protocol, and study metadata
2. Generate TFL shells, SAP, eCRF from the protocol and study design
3. Generate CDISC-compliant SDTM datasets
4. Generate ADaM datasets and analysis-ready variables
5. Produce tables, figures, listings, and traceable statistical programs
6. Run QC checks and generate FDA-ready outputs

We are the first to actually completely automate workflows across SDTM, ADaM, TFL generation, Pinnacle 21 checks, and specification creation.

For ongoing studies, we have compressed work that normally takes months into days.

Why Now

Clinical trials are becoming more complex, but the tools have not improved in the last 20 years (something an actual customer said). At the same time, AI models can finally reason across protocols, SAPs, data dictionaries, clinical datasets, and regulatory standards.

That unlocks something new: biometrics workflows that are not just assisted by software, but executed by agents.

Astraea is not replacing scientific judgment. We are automating the manual programming and data wrangling that slows teams down, while keeping humans in the loop for review, validation, and regulatory decisions.

Our Team

**Joshua Wang: Co-founder, CEO
** Stanford, math/CS. Built multi-agent AI systems and clinical data automation workflows. Previously worked on production AI systems for enterprise customers.

**Sanmay Sarada: Co-founder, CTO
** Stanford CS. Built clinical and hospital data infrastructure across diagnostic software systems. Focused on regulated healthcare workflows and data pipelines.

We started Astraea because people close to us in pharma kept describing the same bottleneck: clinical trial analysis was too manual, too slow, and too dependent on fragmented CRO workflows.

So we built the automation layer we wished existed.

Find us at founders@tryastraea.com

https://tryastraea.com/